_target_: text_recognizer.networks.convnext.ConvNext
dim: 16
dim_mults: [2, 4, 8]
depths: [3, 3, 6]
downsampling_factors: [[2, 2], [2, 2], [2, 2]]